Cough detection method based on data uncertainty learning

Abstract

The invention discloses a cough detection method based on data uncertainty learning. The problem that in the prior art, cough detection accuracy is low in a real environment is mainly solved. According to the implementation scheme, voice data are selected from different public data sets and preprocessed, and the voice data are divided into a training set and a test set; constructing a detector network formed by cascading a noise generation module, a Mel map generation module, a feature prediction module, a mean value and variance prediction module and a full connection module in sequence; setting an objective function of the detector network; setting a learning rate and the maximum number of iterations, and updating the target function through the training set by adopting a stochastic gradient descent method to obtain a trained detector network; and inputting the test data set into the trained detector network to obtain a cough detection result. The method not only can obtain high accuracy in a noiseless environment, but also has better performance under the condition of simulating real noise, and can be used for intelligent detection of cough sound and collection of cough samples.

Description

technical field [0001] The invention belongs to the technical field of voice signal processing, and further relates to a cough detection method, which can be used for intelligent detection of cough sounds and collection of cough samples. Background technique [0002] Cough is the human body’s response mechanism to abnormalities in the respiratory system. It is used to remove pathogens, mucus or foreign bodies. When the respiratory center is in the center, it will cause a cough reflex, which is a protective reflex that can help remove hidden secretions and harmful substances in the respiratory tract, and is beneficial to the human body under normal circumstances. But when frequent, severe and persistent cough occurs, it becomes a pathological state, and its frequency, intensity, and time of occurrence can provide important information for doctors to diagnose clinical patients. Cough detection is the primary part of cough data collection. The quantitative evaluation of cough ...

AI Technical Summary

Problems solved by technology

[0007] One is that when the test set and the training set are not in the same data set, the detection accuracy is low;

[0008] The second is that when there is a complex type and high intensity noise background in the test set, the accuracy rate will be severely damaged;

[0009] The third is that when testing in a real environment, it is difficult to achieve good detection results due to the limitation of training data.

Images